Fare values indicated in the article are accurate as of 28 Dec 2024, unless stated otherwise.

Yesterday marked yet another day where a new fare increase takes effect, with blanket increases of 10 cents for adult card fares, and 4 cents for concession fares. An unfortunate day indeed. It is on this occasion however, when everyone’s mind is preoccupied on the new fare increases, to talk about farebox recovery, a metric often used to measure service performance, especially in Singapore. 

We all know about how LTA’s policy in planning buses is characterised by a certain air of profit-seeking, where valuable connections are severed in the name of judiciously using finite resources, or in other words to make their financial statements look better. Hell, it’s probably in high gear under the Chee Hong Tat MOT administration, with the man himself repeatedly pissing off the public with his incredibly tone-deaf bureaucrat comments despite having been in office for less than a full year! In what way does this penny-pinching attitude manifest? 

Enter the world of farebox recovery, where routes are evaluated based on how well they recuperate their operating costs. Also known as the fare recovery ratio (FRR), it’s often the deciding factor in which routes survive — or die, when LTA and similar transit planning agencies worldwide decide to cull the network. The formula for calculating FRR is very simple… (or maybe not, as you’ll see later):

FRR = fare revenue / operating cost

Farebox recovery is expressed as a ratio of earned fare revenue out of total operating expenditure. For instance, a FRR of 60% means that for every $1 spent operating the service, only $0.60 is earned through fares, with the remainder having to be made up for through subsidies or some other means. Thus a FRR of 1 (100%) implies breaking even, with higher values indicating profits, and lower values indicating losses. 

It may surprise you, but using FRR as a metric to make route planning decisions is a lot more complex than being the proxy for ridership as many would imagine. Like with a lot of things, they’re fine for the job assigned, until they’re the only thing around. As you’ll see shortly, not every cent of fare revenue is made equal, and not every public transport user matters equally, under a FRR-driven analysis framework. 

Dollar votes

In microeconomics, the cheesy way used to refer to an individual’s purchasing power is also known as their dollar vote. Figuratively, the more money you have, the more you’re able to influence the market to produce a certain type of goods, and collectively these form what’s known as price signals. Here’s the kicker — when public transport performance begins to be analysed using FRR, we’re doing precisely this to every rider. To what extent a rider deserves connectivity is determined by on how much they pay. Remember, fare revenue, meaning higher-subsidised demographics are less well-represented at the table, with the needs of full fare-paying white-collar adults taking precedence over them. 

In more concrete terms, a white-collar adult whose maximum payable fare is $2.47, is more than 3 times as important as a student, whose maximum fare is a mere 74 cents! This explains why pleas from students for better connectivity to their schools tend to fall on deaf ears — it’s simply unprofitable to serve students! To attain the same FRR as an identical route serving an adult population under our current fare structures, one would need to transport three times as many students on each bus, while keeping all other factors (ie runtime, vehicle type etc) constant, and without “cheating”, as you’ll see later. From this, it’s quite obvious that it’s extraordinarily difficult for bus services primarily serving more-subsidised demographics to attain a sufficiently high FRR for them to be kept around. Additionally, unlike adult fares which continuously increase until the 40km band, student (and senior) fares are capped at the 8.2km band, which further reduces the amount that they’re paying, and hence further puts services built on student and senior demand at a disadvantage when LTA conducts their reviews.

Two stories come to mind here. Service 902, the two-stop one-way shuttle bus for RP students, was allegedly only introduced by SMRT after the large crowds of students caused severe overcrowding on 169, an industrial long feeder plying in the same direction, causing complaints by working adults riding 169 to their workplaces in Woodlands Industrial Park! In effect, 902 was born out of the frustration of adult riders on 169, and not because of the RP students themselves! In an alternate reality where Woodlands didn’t have an industrial park in the North Coast area, SMRT and LTA would never have bothered to introduce 902!

The other one hits slightly closer to home for more people. Service 852, which links the Bukit Timah school belt to Bukit Batok, Ang Mo Kio and Yishun, has been rumoured to be on LTA’s internal hit list for a long while. Unlike previous rationalisation exercises in the past decade, the one targeting 852 won’t be linked to any MRT line’s opening. Heck, this probably will come as a shock to many, because 852 is regularly well-used, with nearly every trip filling up all seats at some point, as opposed to some other routes which can’t even fully fill a single-deck bus, yet are hardly ever considered for irrationalisation. Why? The large majority of 852’s rider profile, as hinted earlier, comprises students using 852’s (relatively tighter) schedules to get to school much faster, or retirees using it to cross from the west to the north directly. In dollar vote terms, the very brief surge of peak adult commuters swamping 852 in the evening probably produce the same fare revenue as all the students and seniors riding it throughout the whole day!

Most routes with FRRs in the extreme low end tend towards this category — it’s not that their ridership is poor, it’s just that the demographics they serve pay lower fares, and unless said route is the sole bus connection in particular area (re Jurong Industrial services), FRR-based analysis is telling LTA to prepare the axe to fall on said routes. While specific routes cannot be stated here, this perspective of rider demographic pushing the FRR of specific bus routes below a certain threshold is able to explain some rumours around speculated future bus cuts underway. Who knew, bus services benefitting students, seniors and lower-income workers are the first in line for the chopping board!

(By the way, a compound effect manifests here — the abovementioned demographics that pay reduced fares tend also to purchase monthly passes, whose sales revenue are excluded altogether from individual route FRR calculations. This means not only is the average fare paid per rider much lower, there’s also far more riders who are considered to be “riding for free”, adding to ridership counts, but not fare revenue counts!)

The runners who struggle

Look at what’s perhaps Singapore’s most famously overcrowded bus services, exacerbated by limitations imposed by Changi Airport, bus 858, incorporating a long express sector from Yishun to Changi Airport via TPE.

With failboarding being a common occurence on this service due to its incredibly high demand (that even chokes up the hard-carrier of large crowds, bendy buses), what FRR would one expect it to be? Above 100%, being part of the rare few who actually break even and earn profits?

(Video below is a buffer before revealing answer)

Surprise surprise, these popular routes don’t even get anywhere near breaking even, and is a big money loser in LTA’s eyes. It’s definitely nowhere near the bottom because of how many people are riding it, but despite the sheer crowd numbers it’s counterintuively nowhere to be found amongst the most “profitable” bus routes. How, you might ask? 

It helps to learn a bit about the current fare structure. As mentioned in the previous post, public transport fares follow a logarithmic pattern, increasing at a decreasing rate for every additional unit of distance travelled. On top of that, the base fare, which is that large initial fare paid before the gradual increments for subsequent distances travelled, is also credited to the first leg of the journey. The latter bit only matters later, but the former means that a greater determinant of FRR is often total ridership, rather than vehicle occupancy, or distance travelled! 

In fact, this isn’t a uniquely Singapore problem — many transit systems elsewhere often have this distorting effect much more pronounced due to a flat fare system in place, where fare revenue correlates perfectly to ridership count. The problem remains — for as long as fares are not perfectly linear to distance travelled, FRR analysis will disproportionately overrepresent short-distance travellers, at the expense of those travelling longer distances.

Recall the types of demand cycles and their impact on total ridership. You might remember a previous post here stating how empty looking buses can carry more riders than a crush-loaded one, depending on the length and intensity of demand cycles throughout the route. In fare terms, a route’s FRR is decided to a larger extent by its demand cycles too (because fare revenue is tied to total ridership). Thus, FRR-based analysis is unfairly biased against long trunk routes with cumulative demand. Even if these routes form important connections that are heavily relied on by lots of people, many of them are big money losers in LTA’s eyes. Again, no specific numbers here but the large majority of expressway trunk buses, which have predominantly cumulative demand patterns, do horrendously badly at the farebox despite their apparently high ridership. Conversely, most routes that profit exhibit a highly cyclical demand pattern, with many of them being long feeders or feeders with multi-directional demand. 

In that respect, 858 is fortunate to incorporate a long feeder segment between Woodlands and Yishun, and it’s because of it that its FRR is more respectable than many of its long-haul expressway trunk peers. Elsewhere, a worrying number of long expressway trunks are either near, or below LTA’s FRR threshold for irrationalisation considerations. (If publicised, the list of bus services located in this FRR band will shock the public… but it’s with these mechanics that the money-chasing parties within justify eviscerating the network with.) 

The logic is simple: Because each new passenger represents a higher chance for a new base fare to be paid to a given route, shorter demand cycles mean the same limited space on the bus can fetch more potential “$1.19”s, on top of the distance-based fares that apply equally for each passenger demographic. Thus, with the same operating cost for a given route, a cyclical demand profile means a much higher fare revenue, thus in a framework driven by FRR evaluation, unfairly favoured over a more significant long-distance connection. 

It’s also why LTA has been aggressively pushing the hub-and-spoke approach in recent years — besides the irrational belief in perpetual rail reliability, FRR analysis also played a part in them concluding that such was the way forward for our bus network in future. Unfortunately, that’s just an incredibly myopic view of public transport financing, because that drives riders away, shrinking the revenue pie and hurting FRR for their highly-prized rail network. Also, the largely uni-directional demand patterns on the great majority of feeder routes in Singapore would mean the average FRR of all routes doesn’t significantly increase anyway — this inefficiency alone prevents the goal of improving the FRR book values from being reached in the first place! 

Contrary to the claim that 167 has no ridership, it used to enjoy strong demand (SMRT tried to run it on a 1-hour bus frequency right before the handover to Tower Transit, with extremely disastrous results), both before and after the TEL opened! Yet, 167 was considered for a full withdrawal because of its insufficient FRR, arising from the largely cumulative demand from passengers riding all the way from Sembawang to the city! (And to demonstrate the power of repressed demand, 167’s FRR barely budged after slashing service to a half-hourly frequency.) 

Thats why it’s so important to learn about how demand cycles work — the in-built biases inherent between cyclical and cumulative demand routes spills over into analysis of financial viability, which then means certain demand patterns get favoured over others. In the FRR-driven view, the white-collar worker paying $1.09 for a 2km ride to the MRT on a feeder bus is more important than the undergraduate who pays $2.17 for travelling 25km across Singapore on expressway buses. Both have important things to do, but when fare revenue is all that matters (with the assumption that “if the route is gone, just use the other route in the area, regardless whether it actually gets you there”), it really does not favour the less wealthy Singaporean who isn’t able to move to a prime location close to work or school! 

First past the post

Besides having shorter demand cycles, feeders are disproportionately favoured in FRR analysis for one big reason: they get the lion’s share of a passenger’s fare. Here’s the thing — when a journey is made across multiple trips involving a connection, the fare is distributed between each leg based on the additional fare paid along that given sector. Because of the logarithmic fare formula, bus services which are the “first point of contact” of the network tend to have a massive FRR advantage over services that are used as subsequent legs of public transport journeys. Let me illustrate this with an example:

Suppose I start from Admiralty Park (in northern Woodlands), and travel to IKEA Tampines. If I walk to Woodlands ITH and take 168 directly, I pay $2.23, all of which goes to the 23km sector of 168 I ride. However, if I board feeder 903 to Woodlands ITH, then transfer to 168 there, things become a lot different. Of the total $2.26 paid, the 1km sector of 903 receives $1.19 (base fare value, which feeder fares are capped at), while the remaining 23km of 168 receives only $1.07!

Not only does a journey 10 times as far cost only twice as much (which shows the bias any FRR-based analysis would show towards shorter demand cycles), the introduction of a feeder segment before the trunk journey siphons away more than half the fare revenue the trunk would have had! In fact, based on the current fare structure, you would have to travel 29km* on the non-feeder segment of your journey for the feeder portion of your fare (a segment typically less than 5km) to be less than half of your total fare! This corresponds to the 32.3km fare band, which is the first band which has a total fare higher than twice the base fare. 

*This distance is 17km for WTCS holders, and undefined for seniors and students, because of very low fare caps for the latter two concession groups. 

It’s no surprise thus, that feeders (especially those from more HnS-oriented towns, where many residents are only served by a feeder) tend to occupy the upper echelons of the FRR rank list. In general, routes that act as the sole connection to sufficiently large masses of residents tend to have good FRR outcomes too, against expectations of their route type. 

Cheating. 

Despite all the above explanations of how FRR performance metrics work, there’s one last major anomaly that results in what appears to be inexplicable route planning choices. Among the top five routes with highest FRR in Singapore, only one of them logically belongs up there, based on the above deductions of FRR calculation, and it’s the one in 5th place. Of the remaining four, two are feeders too short to amass sufficient demand to even fill up buses for most of the day, and one of them is an express, with a strongly cumulative demand pattern. These are route profiles which, given normal operating logic, would not attain high FRR, much less top the list nationwide. 

The last one deserves special explicit mention, simply because of just how absurd it is for such a route to even have such a high FRR value, much less top the list ahead of the numerous other possible contenders for high FRR! In fourth place, Woodlands shuttle service 902 allegedly has a FRR of 200%. You read that right: that two-stop shuttle to Republic Polytechnic earns twice as much as its operating expenditure, despite running for just 90 minutes every weekday during school term, and with buses not even full! 

To give you a sense of just how logic-defying this is, everything about 902 runs against the conditions needed for strong fare revenue. For one, its main rider base is overwhelmingly students from RP, whose 74 cents have been almost entirely exhausted on the MRT or trunk bus ride to Woodlands, or with what little left from the rare few who commute from areas very near Woodlands. In the context of the morning commute which 902 serves, 902 is also the last trip made for a public transport journey to RP — logically, only RP students use this service, after all. Lastly, its operating cost relative to fare revenue is expected to be high, due to disproportionately massive deadhead distances covered — buses on 902 only operates from Woodlands ITH to RP, before deadheading back to Woodlands ITH to start another round. All the time spent running the bus constitutes operating cost, yet the actual length of time which a bus on 902 carries passengers, is at most 30 minutes a day! 

Based on Datamall statistics for 902’s ridership, even if we assume every rider of 902 pays the adult base fare of $1.09 (these calculations were done prior to the fare increase this year) on 902, the operating cost if FRR is 200% should be $90 a day in total. 902 operates with seven buses, running in a cycle, which means that based on this logic, even if we throw every other operating cost out of the window, each bus captain driving a 902 bus takes home… a measly $12 per day???? A restaurant waiter earns more than that in an hour! Either Datamall statistics are bunk, or that LTA’s FRR statistics are bunk, and the most plausible explanation is that both of the above statements are true.

With incredible statistical anomalies like 902 and many others (especially near the top end of the FRR spectrum), my educated guess for how these absurd results were obtained is that some manipulation was involved to twist the numbers for fare revenue and operating cost respectively, such that routes that otherwise were losing money in the real world, appear to profit massively on paper. 

At this juncture, it’s important to bring up the concept of interlining, or the practice of placing the same bus on multiple routes in the course of a single duty. For instance, some buses on 159 are assigned to cross over to CDS 660/M during the morning and evening peak hours. Peak express routes (such as City Directs and the numerous expresses that function similarly to them) all interline with a trunk or feeder service, to allow the bus to be used on other parts of the bus network outside the peak express period. Some operators, such as TTS and GAS, are also more liberal with interlining various feeder services to enable more fine-tuned allocation of resources across different routes at specific times. Unfortunately, because of a mix of bureaucratic paperwork and interlining practices, FRR calculation becomes much more complicated for quite a number of bus routes here. 

The most likely reason why four of the top five bus services by FRR are there despite not logically belonging there, is most likely the result of manipulating a divorce between operating cost and fare revenue. For vehicle duty blocks that cover more than one route, the bus operator typically classifies the block under one of the services being operated, the “anchor” service, which is usually the one which has the most revenue mileage in the entire block (exceptions do exist). In the above example, the vehicle block crossing between 159 and 660/M would be classified as a “159 duty”, with the anchor being 159. Here’s where the magic happens — for crossover trips (where the bus plies the non-“anchor” service), the operating cost of that trip is very likely to be parked under the anchor route, whilst fare revenue is parked under the actual route it plies. To use the example of the same 159×660 duty above, when the crossover bus is operating a trip on 159, both operating cost and the fare revenue earned during the trip are counted towards 159. No problem here. When it operates on 660 however, the operating cost of the 660 trip is counted towards 159’s total operating cost, while any fare revenue earned on the 660 trip counts towards 660’s total fare revenue! 

In effect, this means that crossover trips artificially reduce the anchor service’s FRR, while increasing their own beyond logical limits. Operating costs on the anchor service are artificially inflated due to the inclusion of the cost of running trips of other services, thus reducing their FRR value. Meanwhile, the interline service, typically one with very few trips to begin with, can eliminate a large part of its operating cost by throwing them elsewhere, while keeping on to the high fare revenue they earn from the brief surge in peak demand they serve. That’s how numerically inexplicable FRR values come about for these services — remember that 902’s 200% FRR value is only in fourth place! This also explains why some routes with very high demand (both cyclical and cumulative) fail to appear on the list of high FRR services, despite logically belonging there — many of them are known to have interlines to peak express or feeder services, which add to their financial burden without the corresponding fare returns. 

outro

If anything, the numerous quirks that distort FRR to paint a vastly different picture of bus service performance from the lived experience riding them is a clear indicator that a revamp to the way bus services are evaluated is necessary. The lowest hanging fruit of them all is to correct for the numerous biases built into the system that discriminate against the travel needs of demographics who are given more subsidies — students and seniors, and the former in particular, also have very valid transport needs that cannot be entirely met with the rail network. Sometimes, it needs to be understood that the bus just does it better, which is why demand for services like 74 and 852 remain very strong, despite consistent nudging from LTA to take the CCL and DTL instead. Arguably, the best way to go about doing this is to start counting people instead of dollars to evaluate route performance — it reflects the purpose of public transport after all, to move people around and get them where they need to be. 

It’s a call to improve upon the datasets that are used to monitor travel demand — after a bit of processing, Datamall’s figures for April 2024 stated that each 902 bus carries only nine passengers per trip, obviously not in line with reality. Currently only aggregate data between pairs of bus stops are provided, which makes it incredibly painful for researchers to pick out demand for bus services for further analysis. The Datamall should be revamped to indicate demand cycles on individual routes per trip (by the hour, as it is done currently), a far more useful visualisation of how people travel — and crowd up our buses as they move through our city. 

More importantly, it’s a call for moderating the financial advisors’ urges, in the name of preserving the integrity of our robust bus network, built up on the backs of numerous valuable trunk routes. The bias against long-haul bus riders is very real, and unfortunately for LTA, those bus riders are here to stay (after decades of rail expansion, expressway and trunk bus services continue to enjoy strong demand while the trains are also well-used). Just because a bus primarily serves long-distance travellers and has other bus routes running alongside it, doesn’t mean it should be cut in the name of “insufficient FRR”. FRR itself is a highly flawed tool that cannot be the judge in deciding bus service rationalisations — it needs a jury consisting of ridership data and a nice map of the overall public transport network, for rationalisation to really be rational. 

One of the PTC’s favourite quotes used to gaslight the population into being more favourable towards fare increases like the one seen yesterday, is to assert that each ride on public transport is subsidised a full dollar by the government. Well, none of us have the power yet to tell them to back off, but at least, it’s good to know that there’s more to that proverbial dollar than meets the eye.

Join the conversation here, and don’t forget to press the like button! Thanks for reading STC.